Interface Dimensioning

Session 4

Dimensioning voice networks – 2G Example ITU ASP COE Training on “Wireless Broadband” Sami TABBANE

5-8 November 2013 – Nadi (Fiji Islands) 1

CONTENTS

Introduction I.

Erlang law

II.

Dimensioning Process and Measurements

III. Traffic and mobility model IV. BSS dimensioning V. NSS dimensioning

2

CONTENTS

Introduction

3

NETWORK DESIGN GENERAL PROCESS

Input data

Network dimensioning

Site capacity planning

Result Network design

Constraints Cost reduction Target quality of service

4

CONTENTS

I. Erlang Law

5

TRAFFIC MEASUREMENTS

Measured in: • Erlang: Voice service (CS), • Erlang

Bits/second: Data service (CS),

• Bits/second: Data service (PS)

6

ERLANG DEFINITION

System load = number of information units (messages or bits) to carry per unit of time. Two parameters: • λ: Average arrival rate, • T: Transmission duration average. Unit: Erlang (Erl) = channel occupation ratio. Erlang table: Allowing the determination of factors including: • Number of traffic channels traffic (in Erlangs), • Blocking rate.

7

ERLANG B LAW (1)

A E N [ A ]=

N

N! N A 1 + + ... + A 1! N!

• EN: Blocking rate (with loss and without queue) • N: number of resources (channels, machines, processes, …) • A: number of Erlangs or Offered Traffic • A = λT (λ: average number of channels request per unit of time and T: Average occupancy duration of the channel)

Standardized formula: CCITT (Rec. Q87). 8

ERLANG B LAW (2)

N = 4 and A = 2.0 Erlangs, the blocking probability is:

2 E 4[ 2 ] =

4

4! 2

3

4

2!

3!

4!

1+ 2 + 2 + 2 + 2 1!

≈ 9 ,5%

For a capacity of N = 6, the blocking probability is: 4

2 E 6[ 2 ]=

6! 2

3

4

5

6

1+ 2 + 2 + 2 + 2 + 2 + 2 1! 2! 3! 4! 5! 6!

≈1, 2%

9

APPROXIMATION OF ERLANG LAW

The approximation of Erlang law is done by the following formula (P.V. Christensen, 1913):

N = A + k

A

Where: • A is the traffic in Erlang, • 10-k is the blocking ratio • k = -log10(blocking ratio).

10

DEFINITIONS

Traffic in Erlang = Resources Occupation duration/Observation duration 1 Erlang = Occupation duration(D2) of a resource during the observation period (D1). If D2 = 15 minutes and D1 = 60 minutes: Traffic = 0,25 Erlang

11

EXAMPLES (1)

Example 1: Traffic of 0,5 Erlang represents the occupation of 1 resource during 50% of the period, or of 2 resources during 25% of the period. Example 2: Traffic of 4 Erlangs represents the occupation of 4 resources during 100% of the period or of 8 resources during 50% of the period.

12

EXAMPLES (2)

Traffic characteristics for fixed phone subscribers:

• Residential subscribers:

0,01 – 0,04 E,

• Business subscribers:

0,03 – 0,06 E,

• PBX:

0,10 – 0,60 E,

• Telephone box:

0,07 E.

13

DIMENSIONING RESOURCES ON THE RADIO INTERFACE

Erlangs in the radio interface

User communication TCH

Traffic ( voice and data)

Signaling SDCCH

Signaling ( establishment, HO)

14

BUSY HOUR (1)

Network dimensioning: Dimensioning a number of channels is based on the busy hour of the day , Special Events (new year, …) not considered (except special events, i.e., with marketing action)

15

BUSY HOUR (2) % Traffic per hour and per day 10% 9% 8% 7% 6% 5%

Peak time Busy Period

4% 3% 2% 1% 0% 0

1

2

3

4

5

6

7

8

9

10

11

Average traffic distribution– days of the week

• • •

12

13

14

15

16

17

18

19

20

21

22

23

Hour of the days

80% of the network cost can be amortized during busy hour 10 to 15% of network sites carry the bulk of the traffic 50% of traffic is carried by 15% of sites at busy hour .

16

16

EVOLUTION OF TRAFFIC DURING THE WEEK

17

BUSY HOUR (DATA TRAFFIC)

18

EXAMPLES OF TRAFFIC DISTRIBUTION – PROFESSIONAL MMS SERVICES

19

EXAMPLES OF TRAFFIC DISTRIBUTION – RESIDENTIAL MMS SERVICES

20

BLOCKING PROBABILITY

Erlang B law is based on the following hypothesis : Random arrival of calls: Poisson process with an average ratio λ, Duration of calls (holding time) according to an exponential distribution (service ratio m, average duration call = 1/µ ), P(service ≤ x) = 1 – e-µx Infinite number of sources of traffic and homogeneity sources, System statistically balanced, Known system load (A = λ/µ ).

21

OFFERED TRAFFIC AND CARRIED TRAFFIC (1)

Offered traffic (loffered)

Available Resources with blocking ratio of x%

Carried Traffic (= throughput= lcarried)

loffered = Arrival ratio of demands, lcarried = Arrival ratio of treated demands, llost = Arrival ratio of rejected demands. loffered = lcarried + llost = l

λcarried = λ(1 – Bc) λlost = λ.Bc 22

OFFERED TRAFFIC AND CARRIED TRAFFIC (2)

Examples

Number of circuits

Blocking rate

Offered traffic

Carried Traffic (max.)

31

2%

22,827 E

22,37 E

32

2%

23,725

23,25 E

23

ERLANG TABLE EXAMPLE

Service level (blocking rate) Channel

1%

2%

3%

5%

10%

20%

40%

Channel

1 2 3 4 5

,01010 ,15259 ,45549 ,86942 1,3608

,02041 ,22347 ,60221 1,0923 1,6571

,03093 ,28155 ,71513 1,2589 1,8752

,05263 ,38132 ,89940 1,5246 2,2185

.11111 .59543 1.2708 2.0454 2.8811

,25000 1,0000 1,9299 2,9452 4,0104

,66667 2,0000 3,4798 5,0210 6,5955

1 2 3 4 5

6 7 8 9 10

1,9090 2,5009 3,1276 3,7825 4,4612

2,2759 2,9354 3,6271 4,3447 5,0840

2,5431 3,2497 3,9865 4,7479 5,5294

2,9603 3,7378 4,5430 5,3702 6,2157

3.7584 4.6662 5.5971 6.5464 7.5106

5,1086 6,2302 7,3692 8,5217 9,6850

8,1907 9,7998 11,419 13,045 14,677

6 7 8 9 10

11 12 13 14 15

5,1599 5,8760 6,6072 7,3517 8,1080

5,8415 6,6147 7,4015 8,2003 9,0096

6,3280 7,1410 7,9667 8,8035 9,6500

7,0764 7,9501 8,8349 9,7295 10,633

8.4871 9.4740 10.470 11.473 12.484

10,857 12,036 13,222 14,413 15,608

16,314 17,954 19,598 21,243 22,891

11 12 13 14 15

16 17 18 19 20

8,8750 9,6516 10,437 11,230 12,031

9,8284 10,656 11,491 12,333 13,182

10,505 11,368 12,238 13,115 13,997

11,544 12,461 13,385 14,315 15,249

13.500 14.522 15.548 16.579 17.613

16,807 18,010 19,216 20,424 21,635

24,541 26,192 27,498 29,498 31,152

16 17 18 19 20

Channel

1%

2%

3%

5%

10%

20%

40%

Channel 24

Spectrale efficiency

EFFICIENCY EVOLUTION

0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 5 10 15 20 25 30 35 40 Offered traffic (Erlangs)

Results of increasing the efficiency according to the traffic and the number of subscribers with a given QoS (non-linear effect). 25

Blockage probability

ERLANG CURVES

Load (Erlangs) 26

ERLANG C DISTRIBUTION (WITH QUEUE)

Queue used to reduce the problem of blocked call: calls which cannot find free resources are queued Erlang C formula dimensioning the number of resources according to the quality of services (e.g.: waiting time before treatment) and the number of calls in the queue. We define: N = number of serves or resources, A = Offered traffic in Erlangs, j = number of waiting calls in the queue, B = probability of losing a call, case without queue (Erlang B), d = Average duration to process a call . 27

ERLANG C FORMULA (1)

Probability that a waiting request can be served: (Erlang C formula ) N .B C= (1) N − A(1 − B )

Average delay (waiting in the queue): (2)

D=

C ×d N−A

Average number of calls in the queue (size of the queue): (3)

J =

A.C N−A

Probability that a delay w is more to t seconds: −(N − A)t/d

Prob(w > t) = C. (4)e

28

ERLANG C FORMULA (2)

Probability that j calls are queued:  A  N

j

C

(5)

Probability that x servers are busy and j places occupied in the queue: (6)  A  A  p(N + j) = C1−    N  N 

j

29

ERLANG C FORMULA (1) • • • • • •

t = 15 sec., Probability to have a resource in t seconds or less = 95%, 3 000 calls/hour, d = 60 sec., A = 3 000×60/3 600 = 50 erlangs, N?

The determination of N requires the calculation of (1) and (4) under having the good value of Prob(w>t). • • • • • • •

Test with N = 55, B = 0.054 (Erlang B formula), C = 0.388, Probability that the delay exceed 15 sec. = 0.11 > (1 – 0.95) NOK Test with a greater N, N = 56: Prob(w>t) = 0.09, always greater than 0.05 NOK Test with N = 57: Prob(w>t) = 0.083 NOK. Test with N = 58: Prob(w>t) = 0.0496 OK. 30

ERLANG C FORMULA (2)

Determination of the number of places in the queue: 1.

To decide the number calls to reject if no free places in the queue

2.

To find j using formula (5). With A = 50 erlangs, N = 57, B = 0.039 and C = 0.246,

Formula (5) j = (log 0.01 – log C)/(log A – log N) = 24.4 So , j = 25. Finally, the average waiting delay = And the average number of calls in the queue = 1.76.

2.1 sec.

31

UTILIZATION OF THE QUEUES

Gestion des demandes d'accès surdemand l'interface par le BSC Management of radio access byradio the BSC Priority Priorité00 (Emergency calls, ré-établissement call re-establishment) (Appels d'urgence, d'appel)

Priority 1 Priorité (Paging, Handover, ...)

TCH allocation Allocation de TCH

... Priority 7 7 Priorité Timeout Rejet de la demande Rejected 32

CONTENTS

II. Dimensioning Processes and Measurements

33

DIMENSIONING PHASES – GSM NETWORK

Um Interface

Cell Abis Interface

Ater Interface

B, C, D Interfaces A Interface 34

TRAFFIC AND DIMENSIONING EVALUATION ELEMENTS

Traffic load estimation: Parameters to determine: • • • • • •

Average call duration, Call arrival rate, Ratio occupation resources, Penetration ratio, Symmetry, service encoding factor , Transfer throughput.

The rate calls per subscribers depends on: • Hour of the day, • Cost call, • Availability of the equipments. 35

DATA TO ESTIMATE THE TRAFFIC

Most important parameters: • Calls rate (number and duration), • Mobility signaling (location update, handover, …). Elements that can be neglected: • Signaling traffic for managing and maintenance, • Handover intra-cellular, • Data and software update and loading, • Supplementary services activation parameters.

36

EXAMPLE OF SCENARIO

Area of 1500 habitants (0,02 Erlang) Growth: +50% per year, Penetration rate: 35 %. 2000 Visitors at busy hour (0,1 Erlang) Growth: +20% par an, Penetration rate: 80 %. Highway with users (0,2 Erlang) Maximum: 300 cars simultaneous, Penetration rate: 75 %, Growth: 5 %. 37

PROCESS

tc: Traffic growth factor in an area, A: Offered traffic (measured or estimated) in the area, M: Security margin. Additional traffic with security margins: Af = A*(1 + tc)*(1 + M)

38

CONTENTS

III. Traffic and mobility model

39

ELABORATION OF TRAFFIC MODEL

Inputs: • Measurements/experiences in existing systems • Or according to preliminary assumptions. Based on values at busy hour.

40

RELATION WITH OTHER PARAMETERS

Traffic and mobility model

Quality of service

Subscribers behaviour

Cellular design

41

TRAFFIC MODEL FOR TCH LOAD EVALUATION (1)

Outgoing call (Mobile to Fix)

Duration (in sec.)

Respective ratios t1

Outgoing call rate t2

t1. t2

TCH occupation duration (in sec.)

52%

(10s+20s+ 120s)*52%= 78,0 sec.

Success: - Establishment, - Ringing, - Conversation.

10 20 120

80 %

No answer: - Establishment - Ringing.

10 20

10 %

6,5%

Busy: - Establishment.

10

10 %

6,5%

65 %

(10s+20s)* 6,5%= 1,95 sec. 10s*6,5%= 0,65 sec.

42

TRAFFIC MODEL FOR TCH LOAD EVALUATION (2 (2))

Incoming calls (Fix to Mobile) Success: - Ringing, - Conversation. No answer: - paging, - Busy. Total

Duration (in sec.)

Respectiv e ratios T1

5 120

55%

Incoming calls rate t2

Duration (in sec.) occupation of TCH

t1. t2

(5s+120s)* 19,25%= 24,0625 sec.

19,25% 35%

40%

14%

5%

1,75%

30

(average occupation in seconds) of TCH

30s*14%= 4,2 sec. 30s*1,75%= 0,525

- Success calls : 71,25% - Failed calls: 28,75%

105,1875 s with the conversation

43

TRAFFIC MODEL FOR THE SYSTEM DIMENSIONING VMS Number of inbox messages Average number of messages per day

5 000 3

Average duration of initial message

10 seconds

Average duration of a message

30 seconds

Average number of messages retrieval per day Average number of messages retrieval Traffic percentage at busy hour

2 10 seconds 10 %

44

DENSITIES [3GPP]

Environment

Density (subs./km2)

Cell type

Dense urban

180 000

Micro/Pico

Urban

100 000

Macro/Micro

Suburban

10 000

Macro

100

Macro

Rural

45

TRAFFIC AND MOBILITY MODEL

Traffic distribution: traffic matrices

46

TRAFFIC MATRICES (1)

Traffic matrices of are essential to characterize the traffic profile in the network : • First choice traffic: End to End traffic, • Traffic matrix: groups of first choice traffics, • Relation between first choice and internal traffics complex.

47

TRAFFIC MATRICES (2)

Designing a traffic matrix: • Exceptional values are not considered, • UIT-T: Second busy hour in the month, second highest monthly value, • The matrix is considered as an input for the planning process.

48

EQUIPMENTS FOR THE PROCESSING OF THE TRAFFIC Served area per A

Served area per B

A

B

Network

To\From

A

B

A

TrAA

TrBA

…

Served area per C

Served area per D

C

D

C

D

…

…

D 49

CONTENTS

IV. BSS/RAN Dimensioning

50

TCH CHANNELS DIMENSIONING

Traffic rate per subscriber Subscribers number

X QoS ( Blockage rate)

Traffic per cell

Erlang Formula

Number of TCH channels 51

SDCCH DIMENSIONING (1)

SDCCH carries: call establishment messages, location update, SMS. Variant 1. SDCCH channel duration occupancy: Per user: dSDCCH = lc*Tc + lloc*Tloc + lSMS*TSMS Per cell: DSDCCH = dSDCCH*users With: • lc: Calls outgoing/incoming rate

• • • • • •

Tc: Average call establishment duration , lloc: Location update ratio, Tloc: Average Location update duration, lSMS: SMS outgoing / incoming ratio, TSMS: Average SMS sending duration , Users: Average number of subscribers per cell. 52

SDCCH DIMENSIONING (2)

Variant 2. Traffic SDCCH processed through TCH traffic: • TrafficSDCCH = TrafficTCH*(1+XSMS+Yloc)*DurationSDCCH/DurationTCH • With: • XSMS = number of SMS / number of calls, • Yloc = number of LU / number of calls.

53

SDCCH NUMBER PER CELL

Calls rate/ sec

Number of SDCCH slots

LU rate /sec Estimated traffic per cell Average Occupancy duration of SDCCH

Erlang Formula Qos

/8 Number of SDCCH channels

54

AGCH CHANNEL DIMENSIONING

AGCH channels transport messages « Immediate Assignment » (1 per block) for the SDCCH channel allocation. Number of necessary AGCH blocks : N_blocAGCH = users*(tca + tloc + tSMS)*Nbl users = Average number of subscribers/cell, tca: outgoing/incoming calls rate per user, tloc: Location update ratio per user, tSMS: SMS-MT and SMS-MO ratio per user, Nbl: number of AGCH blocs used for channel allocation. 55

PCH CHANNEL DIMENSIONING

1 bloc PCH carries a maximum of: • 4 mobiles identities if TMSI is used, • 2 mobiles identities if IMSI is used. Incoming calls and SMSs notified in all the cells of the mobile LAC, Number of necessary PCH blocs : DPCH = users*(tMT + tSMS_MT)*NClac*Mp*Nbl Where: tMT: Incoming call ratio per user, tSMS_MT: SMS_MT ratio per user, NClac: number of cells in the location area, Mp: number of transmitted paging messages per incoming call (2 to 4), Nbl: number of PCH blocs used for paging. 56

NUMBER OF TRX PER CELL Number of channels TCH

Number of Number of Broadcast SDCCH slots channels

∑ Total number of time slots

/8

Number of TRx

57

TCH AND SDCCH CHANNELS AND NUMBER OF CELLS DIMENSIONING

We consider an area where demads is estimated to 100 000 subscribers with the traffic and mobility traffic (in busy hour): • Incoming number SMS = 0,5, • outgoing Number of SMS = 0,51, • Incoming number calls = 0,75, • outgoing Number of calls= 0,8, • Number of intra-VLR location update = 0,6, • Number of inter-VLR location update = 0,2, • Number of maximum TRX per cell = 6, • Average traffic per subscriber = 25 mE

58

TCH AND SDCCH CHANNELS AND NUMBER OF CELLS DIMENSIONING

We assume that: • Received duration of SMS = 1,3 sec., • Sent duration of SMS = 1,6 sec., • Establishment duration of incoming call= 5 sec., • Establishment duration of outgoing call= 18 sec., • Intra-VLR location update duration = 0,7 sec., • Inter-VLR location update duration = 3,5 sec., • Allocation mode of dedicated channel: OACSU, Blocking rate should not exceed 5% for the TCH channels and 1% for the SDCCH channels. What is the number of TCH and SDCCH channels and the number of served cells in the considered area? 59

BSC DIMENSIONING

Capacity BSC characterized by: • Connection capacity, • Processing capacity (data produced by the BTSs and MSCs). Parameters: • Max_BTS: maximum number of supported and controlled BTSs, • Max_CA: maximum number of calls attempts, • Max_TRX: maximum number of TRX, • Max_Port: maximum number I/O ports, • Max_Sig: maximum number of signalization links. 60

RNC DIMENSIONING

RNC capacity determined by: • Maximum number of cells (1 cell = 1 carrier and one scrambling code), • Maximum number of Node B, • Maximum throughput on the Iub interface. Number of RNC: Max(Num_RNC1, Num_RNC2, Num_RNC3) Where: Num_RNC1 = number of cells/(Max_cell*capacity _ margins1) Num_RNC2 = number of Node B/(Max_NodeB*capacity_margins2) Num_RNC3 = (voice load +Data_load_CS+Data_load_PS)*Num_subs/ (Max_throughput_RNC*capacity_margins3) 61

BSC CONTROLLER CAPACITY

Item

BSC n°1

BSC n°2

BSC n°3

Maximum number of transmitters –receptors full throughput

32

192

448

Maximum number of cells

21

140

255

Maximum number of Abis links (connection BSC – BTS)

6

36

84

Maximum number of A links ( MIC link in A interface)

16

40

72

Traffic capacity (Erlangs)

160

960

1 500

1

2

3

Number of cabinets

62

ABIS ABIS//ATER ATER/A /A INTERFACES

• 1 frame = 8 TS at 16 kb/sec • 4 ITs = 64 kb/sec = 1 PCM TS • 1 radio frame = 2 PCM TS

Abis interface

4 Um TS = 1 PCM TS

Ater interface

4 Um TS = 1 PCM TS

A interface

4 Um TS = 4 PCM TS

63

DIMENSIONING BASED ON OBSERVED CONGESTION (1) (1)

Problematic: Interface and equipments re-dimensioning according to the observed congestion ratio 2 approaches: (a) Using Erlang formula, (b) Assume a repetition call rate in case of congestion

64

DIMENSIONING BASED ON OBSERVED CONGESTION (2) (2)

Hypothesis: Trp = Repetition call rate in case of congestion (1 < Trp < 3, Trp = 1,5 for ex.).

Offered traffic = Throughput /(1 – B/Trp)

Where

B: observed blocking ratio, Throughput: Observed traffic flow

65

CONTENTS

V. NSS dimensioning

66

DIMENSIONING OF MSC

MSC capacity is determined by: • Connection capacity (switching) • Treatment capacity of received data from BSCs, other MSCs, HLR, … Expressed in : • Max_BSC: maximum number of supported and controlled BSC, • Max_CA: maximum number of call attempts , • Max_Sig: maximum number of signaling links, • Max_Port: maximum number of I/O lports. 67

MSC TRUNK DIMENSIONING

IWF BSC1 ...

CT1 ...

HLR

PSTN RTC

CTm

MSC VMS

BSCn EIR

MSC

SMS

68

NUMBER OF INTERINTER-MSCS LINKS

Estimated traffic in busy hour→ blocking ratio → (0,5 %)

Erlang B Formula

Number of traffic channel

Number of Trunks

Non STP MSC: ISUP signaling computation 2 links SS7 supported 75 000 BHCA MSC in STP mode: 2 necessary additional links SS7

69

LINK DIMENSIONING MSC - VMS

Traffic model Number of voice mail boxes Average number of messages per day Average duration of greeting message Average duration of message Average number of messages retrieved per day Average duration of retrieval greeting message Traffic percentage in busy hour

5 000 3 10 sec. 30 sec. 2 10 sec. 10 %

With a blocking ratio of 0,5%, the number of necessary PCM links is equal to 2. We should also predict 2 links SS7. 70

HLR AND VLR DIMENSIONING

Capacity VLR determinated by: • Max_Sub: maximum number of subscribers, • Max_Tra: maximum number of database transactions/sec. HLR capacity determinated by: • Number of subscribers, • Authentication requests, - LU, • Subscriber provisioning (add/delete/update), • SMS (SRI-SM, set message waiting, …).

71

LINK DIMENSIONING MSC - HLR

2 links SS7 at least in general are necessary per HLR to manage 100 000 active subscribers. HLR accesses for: • Sending signaling routing to the MSC for location. • location update. • Sending requests for authentication triples by the VLR. 72

LINK DIMENSIONING MSC – SMS ET MSC - EIR

2 links SS7 per MSC links – SMS and MSC – EIR at least in general are necessary for managing 350 000 subscribers.

73

NUMBER OF SIGNALING LINKS

• A signaling link consists in 64 kb/s (E0). • Load (SS7 standard): Umin = 0,2 and Umax = 0,4. Offered traffic (bp/s) = Rs = 8*[lc*Nc*Lc + lsms*(Lsms*Nsms+Msms) + lloc*Nloc*Lloc + lHO*NHO*LHO] N: number of messages per call, L: Average size of message, l: number of calls/sec., Msms: average size of SMS, c: calls, loc: location update, HO: inter-MSC handover. Number of signaling links: NE0 = Rs / (U*64 kb/s) 74

CPU PROCESSING CAPACITY OF THE MSC

Functions

Utilization rate

Call processing, Network mobility management, SMS management, Radio mobility management, Supplementary Services .

75%

Maintenance, Measurements, Background tasks, …

14 %

Sharing rate between the two types of functions

11 %

75

AVERAGE CPU PROCESSING TIME

Events

Duration/ events

Ratio per subscriber In busy hour

Average duration

Outgoing call

25 msec

0,56

14 msec

Incoming call

35 msec

0,30

10,5 msec

Inter-VLR LU

45 msec

0,10

4,5 msec

IMSI Attach

15 msec

0,20

3 msec

... Average consumption per subscriber in busy hour

32 msec

76

CONCLUSION

Conclusion The network dimensioning allow evaluation/validation of network capacity.

the

It requires a precise evaluation of the load traffic (current and expected). The definition of mobility and traffic model for subscriber allow dimensioning different interfaces and network equipment.

77

Thank you

78